What is Schema Markup and why it matters for SEO, GEO, and AEO
Schema markup is structured code — JSON-LD, Microdata, or RDFa — that describes entities and relationships in a standardized format. JSON-LD is Google's preferred format because it sits in the document head (the <head> tag) without touching the page's visible HTML.
The difference between SEO, GEO, and AEO is in the targeted output, not the underlying infrastructure: SEO (search engine optimization) targets rendered links in classic SERPs; GEO (generative engine optimization) targets citations in AI-generated answers — ChatGPT, Perplexity, Google AI Overviews; AEO (answer engine optimization) targets featured snippets and quick answers. Schema markup works as shared infrastructure for all three — it helps Google Search produce rich results and helps AI systems extract and cite information correctly. Google confirms there is no special markup or dedicated AI file (e.g. llms.txt) required for AI Overviews.
How Schema Markup works for classic search engines (SEO)
Google uses schema markup to produce rich results in the SERP: rating stars (Product + AggregateRating), breadcrumbs (BreadcrumbList), recipe cards (Recipe), event listings (Event), job postings (JobPosting). Schema does NOT guarantee rich results — Google decides eligibility based on content quality, guideline compliance, and query relevance.
Priority schema types for classic SEO: Article (blog posts), Product (e-commerce product pages), Review + AggregateRating, BreadcrumbList (site navigation), Organization (homepage), Person (author pages), HowTo (step-by-step tutorials), FAQ (rich result pulled in May 2026, but visible questions remain useful). Validation happens through Google's Rich Results Test and monitoring in Search Console > Enhancements. Structured data must match the visible text — mismatches erode trust and can trigger manual actions.
Schema Markup for AI Overviews and Generative Engine Optimization (GEO)
Google officially confirms there are no additional requirements for AI Overviews beyond standard SEO — markup that's well-implemented for SEO also works for GEO. The schema types that matter most for AI citations: Article (with author, datePublished, dateModified for E-E-A-T), Organization (with logo and sameAs to social profiles, for entity clarity and brand consistency), Person (with name, jobTitle, sameAs for author attribution), HowTo (structured tutorials).
Entity clarity and brand consistency matter more than schema alone: the brand name has to match across every source — schema, visible text, social profiles — for AI systems to recognize and correctly attribute content. See below what the real data shows about the citation lift schema markup alone actually produces.
SEO vs GEO vs AEO, compared
| Dimension | SEO | GEO | AEO |
|---|---|---|---|
| Targeted output | Rendered links in classic SERP | Attributed citations in AI-generated answers | Featured snippet, position zero |
| Goal | Rich results (stars, breadcrumbs, recipe cards) | Citation + attribution in ChatGPT, Perplexity, AI Overviews | Winning the featured snippet |
| Priority schema types | Product, Review, BreadcrumbList, Article, Organization, Person, HowTo | Article (author + dates), Organization, Person, HowTo | Article, HowTo, FAQ |
| Success metric | Impressions, clicks, CTR (Search Console) | AI-engine citation rate (measured manually or with tools like BrightEdge) | Featured snippet win rate |
| Failure mode | Schema errors in Search Console; schema/visible-text mismatches | Missing bylines, inconsistent brand mentions, generic content | Answers too long or too vague for a snippet |
| Crawl requirements | Identical for all three: permissive robots.txt, XML sitemap, Google indexing | ||
| Measurement tools | Search Console, Google Rich Results Test | Manual search in ChatGPT/Perplexity/Gemini, tools like BrightEdge | Search Console (featured snippet tracking) |
| Content format | Structured data + consistent visible text | Answer-first + author attribution + brand mentions | Direct answer, 40-60 words |
| Authority signals | Backlinks, domain authority | Brand mentions, off-site mentions (Reddit, Quora, forums) | E-E-A-T (expertise, experience, authoritativeness, trust) |
Validation and QA: an error-free schema checklist
Validation tools: Google Rich Results Test (search.google.com/test/rich-results) — tests rich-result eligibility. Schema Markup Validator (validator.schema.org) — validates JSON-LD against the Schema.org specification. Search Console > Enhancements — monitors schema errors and warnings Google has detected.
Common errors and how to fix them:
- Missing required field — add the required fields (headline, image, datePublished for Article; name, offers for Product).
- Invalid date format — use ISO 8601 (YYYY-MM-DDTHH:MM:SS+TZ).
- Inaccessible image URL — check the image is public and not blocked by robots.txt.
- Mismatch between schema and visible text — the headline must be the article's actual title, author name the name shown on the page.
- Duplicate schema — avoid emitting the same Article node twice on one page.
Five-step validation checklist: validate JSON-LD with the Schema Markup Validator, test with Google's Rich Results Test, check that schema matches visible text, monitor the Enhancements report in Search Console, and re-validate after every content or schema update.
Integrating schema with popular CMSes: WordPress, Shopify, custom code
WordPress
Yoast SEO auto-generates Article schema for posts and BlogPosting for blog entries, but Organization and Person need manual setup (Settings > Schema > Organization/Person). Rank Math offers a visual schema builder right in the editor for Article, Product, Review, HowTo, FAQ, Event, Recipe — more flexible than Yoast for advanced types. Schema Pro (premium plugin) adds types like LocalBusiness, Service, Course, VideoObject, with conditional logic (schema on specific pages or categories only).
Shopify
Product schema is native: Shopify auto-generates name, image, description, sku, brand, and offers for every product page. Organization schema needs manual setup in theme.liquid (Edit code > Layout > theme.liquid, JSON-LD in <head>). Review schema — manual, or via apps like Loox, Judge.me, Stamped.io, which generate AggregateRating from customer reviews.
Custom code / custom CMS
Inject JSON-LD directly into <head>, via template or plugin, before </head> or before </body> — validate first that it's syntactically correct and matches the visible text. WordPress example, in functions.php, on the wp_head hook:
function add_organization_schema() {
$schema = array(
'@context' => 'https://schema.org',
'@type' => 'Organization',
'name' => 'Site Name',
'url' => home_url(),
'logo' => get_template_directory_uri() . '/logo.png',
'sameAs' => array(
'https://www.facebook.com/pagename',
'https://www.linkedin.com/company/companyname',
),
);
echo '<script type="application/ld+json">' . json_encode($schema) . '</script>';
}
add_action('wp_head', 'add_organization_schema');Impact on Core Web Vitals and page speed
JSON-LD placed in <head> doesn't block rendering — it's a <script type="application/ld+json">, so it's non-blocking and doesn't directly affect LCP. Inline Microdata or RDFa can affect CLS if elements carrying schema attributes lack fixed dimensions. A typical JSON-LD block weighs 2-5KB per page (Article ~2KB, Product with Review ~4KB, Organization + Person ~1KB) — negligible LCP impact if total HTML is under 100KB.
Best practices: minify JSON-LD (no whitespace or newlines), avoid redundant schema (don't emit the same node twice), keep it in <head> as JSON-LD rather than inline Microdata, and monitor HTML size in PageSpeed Insights — target under 100KB total, ideally under 50KB on mobile.
What changed: the FAQPage deprecation timeline
Google pulled the FAQPage rich result from Google Search on May 7, 2026 — the feature (stars, question dropdown) no longer appears in the SERP. FAQPage documentation was removed from Google Search Central in June 2026, and the corresponding data disappeared from the Search Console API in August 2026.
What didn't change: FAQPage remains a valid Schema.org type. The rollback only affected rich-result eligibility in Google Search — the markup itself produces no validation error, and the visible questions and answers remain useful both to readers and to extraction by AI systems (ChatGPT, Perplexity, and Gemini frequently cite FAQ sections). We use FAQPage on dozens of pages across airank.ro, for exactly this reason — its usefulness for AI citation didn't disappear along with the Search rich result.
Speakable schema remains, as of this writing, an uncertain bet: a handful of GEO tool vendors mention it as a priority, but Google hasn't confirmed it as a requirement or an active priority. Recommendation: nice-to-have, not essential — add it only if you want to experiment.
Myths and facts about Schema Markup
Myth 1: Schema markup guarantees citations in AI answers
Fact: an Ahrefs study across 1,885 pages found no meaningful AI-citation lift from adding schema markup alone — schema is entity-clarity hygiene, not a magic lever. Full data below.
Myth 2: You need special AI files (e.g. llms.txt) for AI Overviews
Fact: Google confirms there are no additional requirements for AI Overviews beyond standard SEO — optimizing for AI Overviews is optimizing for the search experience, which is still SEO.
Myth 3: FAQPage schema is essential for GEO
Fact: Google pulled the FAQPage rich result in May 2026 — the markup no longer produces anything in Google Search, but visible questions remain useful to readers and to AI extraction.
Myth 4: Speakable schema is a GEO priority
Fact: not backed by Google documentation as a requirement or priority. A few market tools mention it, but Google doesn't confirm it. Nice-to-have, not essential.
Myth 5: Schema markup hurts page speed and Core Web Vitals
Fact: JSON-LD in <head> doesn't block rendering and has negligible LCP impact if total HTML is under 100KB — a typical block weighs 2-5KB.
What the data shows: the Ahrefs study
The most solid public study on this comes from Ahrefs (Louise Linehan, Xibeijia Guan): 1,885 pages that added JSON-LD between August 2025 and March 2026, compared against a control group of 4,000 pages with no new schema, in a 30-day window before and after. The result: citations rose +2.2% in ChatGPT, +2.4% in Google AI Mode, and fell 4.6% in AI Overviews. No platform showed a meaningful increase.
The practical takeaway: schema markup is a prerequisite for eligibility, not a guarantee of results. Implement it as technical hygiene — entity clarity, brand consistency, correct structure — then put the real effort into content quality, E-E-A-T, and off-site mentions. Those move the needle on AI citations; schema markup alone doesn't. To see what your site already has, the free AI mini-audit checks, among other things, the presence and quality of structured markup.